Multiscale Forecasting Approach of Property Insurance Income via Wavelet Method

Abstract

A multiscale prediction algorithm was presented in this paper based on the monthly data of property insurance income of Zhejiang Province from February 2013 to May 2018, and the correlation of the wavelet coefficients for property insurance income was also discussed at different time scales. By ADF tests, the stability of nonlinear approximation data and detail data were given. Furthermore, the prediction algorithm of nonlinear trend and detail data was studied according to the wavelet decomposition formula and the autoregressive moving average (ARMA) model. Finally, the short-term prediction results from June 2018 to January 2019 were given by the wavelet reconstruction formula. Compared to the prediction results of the traditional ARMA model, the relative error of the prediction results in this paper is smaller and better.